Skip main navigation

Genome-Wide Association Study Identifies a Novel Genetic Risk Factor for Recurrent Venous Thrombosis

Originally published Genomic and Precision Medicine. 2018;11:e001827



    Genetic risk factors for a first venous thrombosis (VT) seem to have little effect on recurrence risk. Therefore, we aimed specifically to identify novel genetic determinants of recurrent VT. To date, genome-wide association studies are lacking.

    Methods and Results:

    We performed a genome-wide association scan in 1279 patients from the MEGA (Multiple Environmental and Genetic Assessment of risk factors for venous thrombosis) follow-up study; 832 patients who remained recurrence free during a median follow-up time of 6.1 years and 447 recurrent VT patients with a median time-to-recurrence of 2.6 years. We analyzed genotype probabilities of ≈8.6 million variants, imputed to the Genome of the Netherlands project reference panel, with a minor allele frequency ≥1% for an association with recurrent VT. One region exceeded genome-wide significance (P≤5×108), mapping to the well-known factor V Leiden locus. Conditional association analyses on factor V Leiden did not yield any secondary association signals. We also identified 52 suggestive association signals (P<1×10−5) at 17 additional loci. None of these loci were previously implicated in VT risk. Replication analyses for 17 lead variants were performed in 350 patients with recurrent VT and 1866 patients with a single VT event from the MEGA follow-up study, THE-VTE (Thrombophilia, Hypercoagulability and Environmental Risks in Venous Thromboembolism) study, and LETS (Leiden Thrombophilia Study). We observed an association with recurrence for an intergenic variant at 18q22.1 with an odds ratio of 1.7 (95% confidence interval, 1.2–2.6) per copy of the minor allele.


    We confirmed the association of factor V Leiden and identified a novel risk locus at 18q22.1 in the first large genetic study on recurrent VT.


    See Editorial by Voora and Becker

    Clinical Perspective

    Venous thrombosis (VT), a common complex disorder with a strong genetic basis, is associated with considerable mortality and morbidity. In 20% to 30% of the patients, VT recurs within 5 years of the first event. Patients whose first thrombotic event is not precipitated by any of the established clinical risk factors have an increased risk of recurrence. Apart from factor V Leiden, few genetic determinants of recurrence risk are known. In addition, it is unclear whether different genetic variants impact recurrence compared with a first VT, for example, genetic variants that impact the recanalization of the vein after a VT. We, therefore, conducted a genome-wide association analysis for recurrent VT to identify common genetic variants associated with recurrence risk. In addition to factor V Leiden, we observed a genetic locus at 18q22.1 to be significantly associated with recurrence and found additional evidence for this association in a replication study. The associated region may include regulatory sequences that could impact the expression of distant or nearby genes, such as the nearest protein-coding gene TMX3 that is involved in platelet function. Formal replication and further (functional) work are needed to provide insight into the underlying biological mechanism.

    Approximately 20% to 30% of patients with a first venous thrombosis (VT) develop a recurrence within 5 years of the first event,1,2 and therefore predicting and preventing recurrence are of crucial importance. However, risk factors for a first event do not predict recurrence well and hence risk profiling is difficult.36 Recurrence risk is the highest among patients whose thrombotic event was not provoked by transient risk factors, such as surgery and immobilization.1,2,79 In particular, previous studies have shown that patients with a first unprovoked event have a 2- to 3-fold increased risk of recurrence compared with patients with a first provoked event.79 This suggests that patients with recurrent VT are enriched for genetic risk factors. The minor effects of determinants of first events on recurrence on the relative risk scale can be explained by the difference in absolute risks of first and recurrent VT and index event bias.10,11

    In addition, different genetic variants may play a role in recurrence than in first thrombosis, for example, factors that affect clot lysis or the recanalization of the vein after a thrombotic event. To date, few studies have focused on recurrence-specific genetic risk factors. Zee et al12 studied a panel of 86 variants in 56 candidate genes and observed suggestive associations with recurrent VT for 4 variants. In addition, homozygosity of Ser128Arg in the E-selectin gene and length of a GT-dinucleotide repeat in the promoter of the gene encoding heme oxygenase 1 have been linked to recurrent VT in an Austrian study.13,14 However, none of these findings have to date been confirmed in large independent studies.

    To identify novel genetic determinants of recurrent VT, we performed the first genome-wide association study (GWAS) on recurrence in 447 patients with recurrent VT (median time-to-recurrence 2.6 years) and a sample of 832 patients who remained recurrence free during a median follow-up time of 6.1 years in the MEGA (Multiple Environmental and Genetic Assessment of risk factors for VT) follow-up study.15 To validate our findings, we additionally performed a replication study of the newly identified risk variants among 350 patients with recurrent VT and a sample of 1866 patients with a single event only from 3 cohort studies.

    Material and Methods

    For purposes of reproducing the results, requests for access to the data (summary statistics and replication results) and analytic methods will be considered by the authors.

    GWAS Analysis

    Study Population

    We included patients from the MEGA follow-up study, a large population-based cohort study on risk factors for recurrent VT. Details of this study have been described elsewhere.15 In short, 4956 patients with a first deep vein thrombosis (DVT) of the leg or pulmonary embolism, who were enrolled in the MEGA case–control study between 1999 and 2004,16 were invited to participate. Follow-up started at the date of the first event. Between 2008 and 2009, questionnaires related to recurrent VT were sent to the patients. Occurrence of recurrent VT was determined by information from patients, anticoagulation clinics, and treating physicians according to a decision rule.15 Follow-up ended when a recurrent VT occurred, the patient died or migrated, or when the questionnaire was returned, whichever occurred first. For the patients who died, information on the cause of death was retrieved from the national registry of death certificates. If no questionnaire was returned, patients were considered lost to follow-up.

    For the GWAS analysis, 1499 patients were selected according to the following process (flow diagram is shown in Figure I in the Data Supplement). First, patients who had not provided a high-quality blood sample or buccal swap for DNA analysis were excluded (667 of 4956 eligible patients). In addition, we excluded all patients who had been diagnosed with cancer (n=457). We then selected all patients for whom a recurrent VT event was reported at time of sample selection for the current analysis (n=542). Of these, 16 recurrences were classified as uncertain recurrences according to the decision rule,15 and these patients were subsequently analyzed as recurrence-free patients. In addition, we randomly sampled 957 patients, totaling 973 patients who remained without a recurrent event during a median period of 7.1 years (interquartile range [IQR], 5.5–8.4). Follow-up was incomplete for 19.5% of these patients because some died without recurrence (n=11), whereas others were last seen at the anticoagulation clinic (n=77) or at time of blood sampling for the MEGA case–control study (n=102). Patients with incomplete follow-up were followed for a median period of 312 days (range, 60 days to 9.7 years). Because these patients did not or no longer visit the anticoagulation clinic, which monitor anticoagulant treatment, it is unlikely that these patients experienced a recurrent VT and, therefore, these patients were considered as recurrence-free patients in the GWAS analysis. We performed a sensitivity analysis for the top GWAS findings in which we excluded patients with incomplete follow-up and patients who had an uncertain recurrent event.

    This study was approved by the Medial Ethics Committee of the Leiden University Medical Center, and all participants gave written informed consent.

    GWAS Quality Control and Imputation

    Genome-wide genotyping was performed with the Illumina Human660-Quad v.1 BeadChip (Illumina Inc., San Diego, CA) at Centre National de Génotypage (Institut de Génomique, Evry, France). Genotyping was successfully completed for 1461 patients, of whom 1426 had a call rate of at least 98%. Additional exclusions at the individual level included discrepancy between self-reported and genotypic sex, abnormal level of autosomal heterozygosity (false discovery rate <1%), and ethnic outliers based on multidimensional scaling analysis of the identity-by-state matrix. Furthermore, 32 patients withdrew their consent for the MEGA follow-up study, leaving a total of 1279 patients for imputation and association analyses (447 patients with a recurrence during follow-up and 832 recurrence-free patients). The following exclusions were applied to identify a final set of 497 563 high-quality variants: minor allele frequency (MAF) <1%, genotyping call rate <98%, significant deviation from Hardy–Weinberg equilibrium (P<1×10−6) in patients with a first event only. All quality control procedures were performed with the R-package GenABEL.17

    After the conversion of the genomic positions from hg18 to hg19 using the UCSC Genome Browser LiftOver tool, imputation of 19.6 million autosomal variants was performed using IMPUTE2 software18 according to the Genome of the Netherlands reference panel (based on 250 trios; GoNL release 4).19 Before the association analyses, we excluded variants with an MAF <1% or an imputation quality score I <0.5.

    Statistical Analysis

    Imputed genotypes of 8.6 million variants were tested for an association with recurrent VT using SNPTEST version 220 by means of logistic regression with the missing data likelihood score test, which takes the uncertainty of the imputed genotypes into account. All analyses were adjusted for age and sex. We assumed an additive mode of inheritance. The level of genome-wide significance was set at P<5×10−8, whereas the threshold for highly suggestive association signals was set at P<1×10−5. To identify independent secondary association signals at a locus, we performed conditional analyses on the lead variant or the previously reported VT risk variant. In addition, we grouped associated variants in clumps based on linkage disequilibrium and genomic distance according to standard settings in PLINK.21 Regional association plots were created with LocusZoom,22 and functional annotation of the variants was performed with AnnoVar.23

    The quantile–quantile plot of the genome-wide test statistics against the expected null distribution showed no appreciable evidence of inflation because of population stratification or genotyping artefacts (Figure II in the Data Supplement). Likewise, the genomic inflation factor (lambda24) before and after imputation was 1.033 and 1.001, respectively. None of the first 4 principal components were associated with recurrent VT, and these were, therefore, not included as covariates in the association analyses.

    Look-Up of Previously Reported Risk Variants

    To validate previously reported genetic associations with (recurrent) VT that may not have attained genome-wide significance in our study, we specifically explored the association results for 17 variants. Selected variants were either previously shown to be associated with recurrence only1214,25 or reached genome-wide significance in 1 of the 2 recent GWAS studies on first VT.26,27 Effects were calculated per copy of the risk allele based on the reporting in the original studies. Additional information on the selected variants is provided in Table I in the Data Supplement. Two variants (rs3025058 and rs3074372) could not be studied because of the absence of (tagging) variants in the GWAS, and 1 variant (rs114209171) could not be studied because it was located on the X chromosome.

    Replication Analysis

    Study Population

    The replication analysis was conducted in 350 patients with recurrent VT and a sample of 1866 patients with a single event only. These individuals were included from 3 European studies into VT risk, that is, the MEGA follow-up study, the LETS (Leiden Thrombophilia Study) study,4 and the THE-VTE (Thrombophilia, Hypercoagulability and Environmental Risks in Venous Thromboembolism) study.28 From the MEGA follow-up study, we included 155 patients with recurrent VT who had not been included in the original GWAS or who were excluded during the quality control procedures of the GWAS. In addition, we randomly sampled 929 patients with a single VT event only, of whom 72.9% had complete follow-up.

    LETS and THE-VTE study are both population-based case–control studies into risk factors for VT with subsequent follow-up of the patients with VT. The study designs are similar to that of the MEGA study and have been described in detail previously.4,28 In LETS, 474 consecutive patients with a first DVT in the leg or arm were recruited at 3 anticoagulation clinics in or near Leiden. Patients were subsequently followed for recurrence until 2000 using repeated questionnaires. Follow-up started 90 days after the date of the first event and ended at the date of recurrence, date of death, date of emigration, or the end of the study, whichever occurred first.4 A total of 471 patients had a DNA sample available for genotyping. Of these, 90 patients developed a recurrence during a median follow-up of 8.0 years (IQR, 6.8–9.0). Follow-up was complete for 88.2% of the recurrence-free patients. THE-VTE is a 2-center case–control study, in which 796 consecutive patients with a first VT were enrolled in Leiden and Cambridge (United Kingdom).28 Patients were subsequently followed for recurrence starting at the date of the first event. In Leiden, follow-up ended when a recurrent event occurred, when a patient died or migrated, or when patients were untraceable, whichever occurred first. For patients included in Cambridge, recurrence status was checked on July 1, 2013, using hospital records. In the absence of recurrence or death, this date was registered as the end of follow-up. For the current analysis, we excluded patients who did not have a DNA sample available (n=135). During a median follow-up of 5.4 years (IQR, 4.2–6.6), 105 of the 661 patients experienced a recurrent VT event. Follow-up was complete for 88.5% of the patients with a single VT event only. In both LETS and THE-VTE, individuals with a recent cancer diagnosis were not enrolled.

    All participants gave written informed consent. The THE-VTE and LETS study were both approved by the Medical Ethics Committee of the Leiden University Medical Center. In addition, THE-VTE was also approved by the National Health Service Research Ethics Committee in Cambridge, United Kingdom.


    For each novel locus that showed a highly significant association with recurrent VT in the discovery GWAS, we selected the lead variant or the variant with the largest functional impact. These variants were genotyped with predesigned or custom-made TaqMan assays (Life Technologies, Thermo Fisher Scientific) according to manufacturer’s specifications. Primer design failed for 3 variants (rs9834479 in ROBO1, rs61504683 in LPPR3, and rs111750150 in TSPEAR), which were subsequently replaced by variants in high linkage disequilibrium (r2>0.8) in our GWAS study population or based on the CEU 1000 Genomes population using SNAP software.29

    Statistical Analysis

    Association with recurrent VT was assessed using logistic regression analyses adjusting for age, sex, study, and study center in case of THE-VTE. Patients who were lost to follow-up were analyzed as recurrence free. These patients remained without a recurrent event during a median follow-up period of 1.2 years (IQR, 0.7–3.4). To account for multiple hypothesis testing, the threshold for statistical significance was set at 0.05 divided by the number of variants tested in the replication analyses. We also calculated the false discover rate. In addition, we performed a subanalysis including only the patients from LETS and THE-VTE in a Cox regression model to calculate hazard ratios with 95% confidence intervals (95% CI). In this analysis, patients who were lost to follow-up were censored at the last date known to be recurrence free. To ensure comparability of follow-up time between the LETS and THE-VTE study in the Cox regression analysis, we recalculated the follow-up time in THE-VTE to start 90 days after the date of the first event.

    For the variant that replicated, we performed a meta-analysis of the results obtained in the replication cohorts and in the original GWAS to obtain the most robust estimate of its effect size. For this, we used a fixed-effects model based on inverse variance weighting as implemented in the METAL software.30 Heterogeneity was assessed by the Cochran’s Q statistic and the I2 index.

    Discriminative Value

    To explore the potential clinical value of the 2 identified and validated genetic risk loci, we assessed the discriminative accuracy of 2 prediction models: a clinical model and a combined model to which we added dosages of 2 genetic variants (rs6025 and rs9946608). The clinical model included sex, age, event type (DVT only versus pulmonary embolism with or without a DVT), and provoking status (recent surgery, trauma, immobilization, hormone use, pregnancy, and travel). We fitted both models in the GWAS population, which had complete clinical information for 1260 individuals (443 recurrence patients and 817 patients with a first VT only). Areas under the receiver-operating characteristic curves were constructed using the predicted risks derived from logistic regression models. We calculated and compared the areas under the receiver-operating characteristic curves of the 2 prediction models using DeLong test for correlated receiver-operating characteristic curves as implemented in R-package pROC.31


    GWAS Analysis

    Population Characteristics

    After quality control assessments, 447 patients with a recurrent VT and 832 patients with a single VT event were included in the genome-wide association analyses. Overall, these patients had been followed for a median period of 6.1 years (IQR, 2.2–7.9). Seventeen percent of the recurrence-free patients did not complete follow-up because some died without recurrence (n=9) or had an uncertain recurrent event (n=10), whereas others were last seen at the anticoagulation clinic (n=46) or at time of blood sampling for the MEGA case–control study (n=75). The mean age at time of the first event was 48.1 years (SD, 12.9) and 49% of the patients was a man. Sixty-one percent of the patients had a first DVT of the leg, whereas 29% had a pulmonary embolism and 10% of the patients were diagnosed with both. Compared with patients with a single VT event, patients who experienced a recurrence were more often men and had more often a first unprovoked event (Table 1).

    Table 1. Characteristics of GWAS Study Population

    Patients With a First VT Only n=832Patients With a Recurrent VT n=447
    Age at first event, y, mean (SD)47.0 (12.8)50.2 (12.7)
    Male sex, n (%)339 (40.7)287 (64.2)
    Body mass index, kg/m226.8 (4.7)27.1 (4.5)
    Smoking, n (%)297 (35.7)144 (32.9)
    First event was unprovoked,* n (%)248 (29.8)220 (49.2)
    Duration of anticoagulant therapy, median days (IQR)183 (110–213)185 (111–212)
    Type of first event
    DVT, n (%)497 (59.7)283 (63.3)
    PE, n (%)265 (31.9)102 (22.8)
    DVT and PE, n (%)70 (8.4)61 (13.8)

    DVT indicates deep vein thrombosis; GWAS, genome-wide association study; IQR, interquartile range; PE, pulmonary embolism; and VT venous thrombosis.

    *Provoking factors: recent surgery, immobilization (plaster cast, bedridden at home, hospitalization), hormone use, pregnancy or postpartum, and travel.

    Association Analyses

    We assessed the association between 8.6 million variants and recurrent VT. The Manhattan plot of the GWAS results is shown in Figure III in the Data Supplement. Nineteen variants, all mapping to the F5 region, were associated with recurrent VT at genome-wide significance (Table II in the Data Supplement). The lead variant mapped to a noncoding sequence in F5 (rs2213868, MAF, 14%; P=2.67×109). The F5 locus also included the established VT-associated variant factor V (FV) Leiden (rs6025, MAF, 9.6%; P=1.28×108), of which the T-allele was associated with a 2.4-fold increased risk of recurrent VT (95% CI, 1.75–3.15). Conditional analyses on rs6025 did not reveal any secondary association signals at the locus (Figure IV in the Data Supplement). Of the genome-wide significant variants, the lowest remaining P was 0.02 for rs2213868 (Table II in the Data Supplement).

    We additionally identified 52 variants that showed suggestive evidence of an association (P<1.0×105) with recurrent VT (Table III in the Data Supplement). Of these, 9 variants were part of the F5 locus and were no longer associated with recurrent VT when conditioning on FV Leiden. The other 43 variants mapped to 17 loci, mainly at noncoding sequence. None of the variants or gene regions have previously been implicated in the risk of recurrent or a first VT. We did not identify independent association signals at any of these loci when conditioning on the lead variant of each locus (data not shown). The effect estimates of the lead variants did not materially change in a sensitivity analysis excluding patients who were lost to follow-up although CIs became wider because of the smaller sample size (Table IV in the Data Supplement). Likewise, all lead variants remained associated with recurrence risk, with similar effect sizes, in a sensitivity analysis adjusting for provoking status (Table V in the Data Supplement).

    Furthermore, we aimed to replicate previous genetic associations with recurrent VT and to explore associations for variants recently reported in GWAS analyses on first VT. Results are reported in Table 2. We assessed the association of 8 variants that reached genome-wide significance in 2 recent GWAS studies. Besides the association with FV Leiden, we observed a nominal association with recurrent VT for FGG rs2066865 (odds ratio, 1.30; 95% CI, 1.09–1.56) and F5 rs4524 (odds ratio, 1.25; 95% CI, 1.02–1.54). The recently identified risk variants in SCL44A2 and TSPAN15 showed no evidence of an association with the risk of recurrence (rs2288904, odds ratio, 1.14; 95% CI, 0.90–1.44 and rs78707713, odds ratio, 1.14; 95% CI, 0.85–1.54, respectively). In addition, 5 variants that have previously been linked to recurrent VT risk were not associated with recurrence in the present GWAS analysis (Table 2).

    Table 2. GWAS Look-Up of Genetic Variants Previously Associated With First or Recurrent Venous Thrombosis

    Literaturers IDChr.PositionGeneA1/A2EAFInfoOR (95% CI)P Value
    First VTrs45241169511755F5C/T0.8001.001.25 (1.02–1.54)0.032
    rs60251169519049F5C/T0.0960.942.35 (1.75–3.15)1.28×10-8
    rs20668654155525276FGGG/A0.3391.001.30 (1.09–1.56)0.003
    rs42534174187199005F11T/C0.4830.951.17 (0.99–1.39)0.068
    rs5295659136149500ABOT/C0.4471.001.19 (1.00–1.42)0.055
    rs787077131071245276TSPAN15C/T0.9150.981.14 (0.85–1.54)0.381
    rs17999631146761055F2G/A0.0210.771.25 (0.64–2.42)0.516
    rs22889041910742170SLC44A2A/G0.8391.001.14 (0.90–1.44)0.265
    rs60876852033777612PROCRG/C0.3810.981.05 (0.89–1.25)0.543
    Recurrencers53611169701060SELET/G0.1211.001.14 (0.89–1.47)0.296
    rs1799864346399208CCR5G/A0.0670.830.91 (0.63–1.32)0.622
    rs805297631622606APOMC/A0.3130.981.07 (0.90–1.28)0.447
    rs662794937446PON1T/C0.2981.000.93 (0.77–1.11)0.405
    rs18007751656995236CETPC/A0.4931.001.03 (0.87–1.22)0.718

    Effects were calculated per copy of the risk allele, as reported in the original study, and adjusted for age and sex assuming an additive mode of inheritance. A1 indicates reference allele; A2, effect allele; Chr, chromosome; CI, confidence interval; EAF, effect allele frequency; GWAS, genome-wide association study; info, imputation quality info score; OR, odds ratio; and VT, venous thrombosis.

    Replication Analyses

    To eliminate false-positive findings, we next performed a replication study in 350 patients with recurrent VT and 1866 patients with a single event only from 3 population-based cohorts. Overall, patients were followed for recurrence for a median period of 6.1 years (IQR, 3.8–7.8) albeit follow-up started at different moments in time (see Material and Methods). Follow-up was complete for 83.7% of the patients.

    For each of the 17 loci, we genotyped either the lead variant or the variant with substantial functional impact and tested these for an association with recurrent VT in the replication cohorts. Results of the replication analyses are presented in Table 3. For 2 variants, rs142454359 and rs117161628, we observed only 1 carrier and, therefore, these variants could not be studied in detail. We observed an association with recurrent VT for 1 variant, whereas the remaining variants showed no evidence of an association with recurrent VT. Variant rs9946608 is located in an intergenic region at 18q22.1 and was associated with a 1.7-fold (95% CI, 1.16–2.59; P=0.008; false discovery rate, 0.136) increased recurrence risk per copy of the minor allele. Similarly, we observed a hazard ratio of 1.69 (95% CI, 1.18–2.42) per copy of the minor allele of rs9946608 for recurrence risk in a subanalysis of patients from the LETS and THE-VTE cohorts. When we meta-analyzed the results obtained in the replication cohorts and the discovery GWAS, the minor allele of rs9946608 was associated with a 2.2-fold increased recurrence risk (Table 4; 95% CI, 1.62–2.98; P=4.83×107). There was no evidence for heterogeneity across the 3 replication cohorts (Q statistic, 1.12; I2, 0.00; P=0.57) nor across the replication cohorts and the discovery GWAS (Q statistic, 3.66; I2, 18.1; P=0.30).

    Table 3. Main Findings of GWAS and Replication Study for Lead Variants at Highly Suggestive Loci of GWAS

    rs IDChrPosition(Nearest) GeneA1/A2InfoGWASReplication
    MAFOR (95% CI)P ValueMAFOR (95% CI)PValue
    rs1123499201159933483LINC01133C/T0.980.0116.90 (3.06–15.6)3.36×1060.0060.21 (0.03–1.55)0.125
    rs1444825392164295170FIGNA/G0.600.0158.11 (3.33–19.8)4.14×1060.0111.85 (0.92–3.71)0.084
    rs34029315310571102ATP2B2A/G0.950.0812.16 (1.57–2.97)2.48×1060.1011.01 (0.76–1.34)0.931
    rs41499647350525154CACNA2D2C/T0.910.2051.61 (1.31–1.99)8.90×1060.2040.81 (0.65–1.01)0.063
    rs6548639*379687975ROBO1C/T0.700.3801.59 (1.29–1.96)1.39×1050.5451.03 (0.87–1.23)0.699
    rs114497105513759735DNAH5C/T0.700.0146.91 (2.97–16.1)7.55×1060.0121.26 (0.60–2.62)0.547
    rs1424543595135637432TRPC7G/A0.580.0364.06 (2.28–7.25)2.07×1060.000∞ (0.00-∞)1.000
    rs794385895158397065EBF1G/T0.800.0224.32 (2.30–8.11)5.21×1060.0180.93 (0.49–1.77)0.820
    rs7806964068859837RP11-314C16.1C/T0.630.0147.69 (3.18–18.6)6.01×1060.0200.86 (0.45–1.64)0.639
    rs23343216110567409METTL24G/A0.970.0840.50 (0.37–0.68)7.86×10−60.0770.82 (0.59–1.15)0.256
    rs142720518939156170CNTNAP3T/C0.690.0740.41 (0.28–0.60)6.11×1060.0650.88 (0.61–1.28)0.515
    rs476698612113076475PTPN11C/T0.850.0572.42 (1.64–3.58)8.75×10−60.0540.74 (0.49–1.13)0.159
    rs99466081865817281RP11-638L3.1T/C0.940.0372.91 (1.83–4.61)5.76×1060.0331.73 (1.16–2.59)0.008
    rs351995*19809732PTBP1C/A0.600.4490.62 (0.50–0.78)2.05×1050.4720.97 (0.82–1.14)0.676
    rs203551201192766C20orf202T/G0.940.2371.61 (1.31–1.97)5.26×1060.2461.00 (0.83–1.21)0.987
    rs785714202136377390RUNX1T/A0.990.0422.58 (1.70–3.93)9.88×1060.0410.79 (0.50–1.25)0.311
    rs117161628*2146138322TSPEARC/TNANANANA0.000∞ (0.00–∞)1.000

    GWAS: effects calculated per copy of the minor allele, adjusted for age and sex.

    Replication: effects calculated per copy of the minor allele using a logistic regression model, adjusted for age, sex, study, and country of study. A1 indicates major allele; A2, minor allele; Chr, chromosome; CI, confidence interval; GWAS, genome-wide association study; info, imputation quality info score, MAF, minor allele frequency; NA, not applicable; and OR, odds ratio.

    *Primer design failed for rs9834479 in ROBO1, rs61504683 in LPPR3, and rs111750150 in TSPEAR, and each of these were replaced by tagging variants. For rs111750150, a tagging variant was selected based on 1000G CEU population using SNAP software 28 because there was no tagging variant available in the GWAS.

    This variant was not the lead variant at this locus, but it was selected on its functional impact (linkage disequilibrium with lead variant r2 0.93).

    Table 4. Association Results of rs9946608 in 3 Replication Cohorts

     Recurrent VT patients1321000.0351.43 (0.71–2.87)
     First VT patients8504310.025Reference
     Recurrent VT patients661010.0782.40 (1.17–4.90)
     First VT patients3302510.038Reference
     Recurrent VT patients95910.0521.61 (0.78–3.29)
     First VT patients5193600.032Reference
    Meta-analysis1.76 (1.17–2.65)
    Combined with GWAS2.20 (1.62–2.98)

    Results were meta-analyzed using a fixed-effect meta-analysis model based on inverse variance weighting. Heterogeneity was assessed by the Cochran’s Q statistic and the I2 index. Across the 3 replication cohorts, the heterogeneity measures were as follows: Q, 1.12; I2, 0.00; P, 0.57. For the 3 replication studies and the discovery GWAS, we observed a Q of 3.66, I2 18.1, and P 0.30. In the GWAS, the MAF of rs9946608 was 0.583 in recurrence patients and 0.256 in patients with a first VT only. CI indicates confidence interval; GWAS, genome-wide association study; LETS, Leiden Thrombophilia Study; MAF, minor allele frequency; MEGA, Multiple Environmental and Genetic Assessment of risk factors for venous thrombosis; OR, odds ratio; THE-VTE, Thrombophilia, Hypercoagulability and Environmental Risks in Venous Thromboembolism; and VT, venous thrombosis.

    We subsequently interrogated several publicly available databases for potential mechanistic information on rs9946608. No significant expression quantitative trait loci have been reported in Genotype-Tissue Expression32 for rs9946608 or any of the linked variants (r2>0.8). We used RegulomeDB,33 which integrates information from the ENCODE34 and Roadmap Epigenomic35 projects, to assess whether rs9946608 or linked variants may have a regulatory function. There is minimal evidence that several variants at this locus, including rs9946608, may affect transcription factor–binding affinity (Regulome score 4). In some cell lines, DNase peaks in the chromatin structure have been identified using DNase sequencing. Genes located nearby, which could be potential target genes, are 2 long intergenic noncoding RNA genes (RPH11-526H11-1 and RP11-638L3.1) and protein-coding gene TMX3. The latter encodes thioredoxin-related transmembrane protein 3, which has been detected in human megakaryocytes, platelets, and at the platelet surface of both resting and stimulated platelets.36

    Discriminative Value

    In a preliminary analysis, we explored the added discriminative value of FV Leiden and rs9946608 to a prediction model with clinical risk factors alone. The areas under the receiver-operating characteristic curve of the clinical prediction model, which included sex, age, event type, and provoking status, was 0.65 (95% CI, 0.61–0.68). Predictive accuracy of recurrence risk significantly improved when adding the 2 genetic risk variants to the model (areas under the receiver-operating characteristic curve, 0.68; 95% CI, 0.65–0.71).


    This GWAS is the first large-scale genetic discovery effort for recurrent VT. Previous studies were either small or focused on candidate gene variants, such as FV Leiden and prothrombin G20210A. The high recurrence rate of VT, especially in patients with a first unprovoked event, and the subsequent lifelong treatment with anticoagulants make it important to uncover the genetic and biological architecture of recurrent VT. Here, we confirm the association of FV Leiden with recurrence and identify a novel potential risk locus at chromosome 18q22.1.

    Genome-wide significance was attained by several variants at the F5 locus, which included the well-known risk variant FV Leiden. We observed a 2.4-fold increased risk of recurrence per copy of the T-allele of FV Leiden, which is slightly higher than previously reported,3,4 albeit still lower than the risk estimates observed for a first VT.26,27 There were no secondary association signals observed at the F5 locus. Known VT risk variant rs4524, which has been shown to affect the risk of a first thrombotic event independent of FV Leiden,26,37 was only nominally associated with recurrent VT. This may suggest that FV Leiden is the key determinant at the F5 locus of recurrence risk.

    We additionally identified 43 variants at 17 novel loci associated with recurrent VT at suggestive significance (P <1.0×105). We sought to replicate these findings in independent samples from 3 studies. Our results suggest that carriers of rs9946608-C have a 1.7-fold increased recurrence risk compared with noncarriers. We observed little evidence for statistical heterogeneity between the replication studies which could explain our findings. Formal replication is needed to confirm the association between rs9946608 and recurrent VT as the meta-analysis of the GWAS and the replication studies did not reach genome-wide significance. From a clinical perspective, it would also be interesting to evaluate whether this variant has a differential effect on recurrent DVT or pulmonary embolism, which was now impossible to study because of low number of patients.

    Variant rs9946608 and proxies map to noncoding sequence at chromosome 18q22.1 and have not been implicated in disease risk before. If the association with recurrence risk is true, this intergenic locus has most likely a regulatory function. We observed some evidence of transcription factor binding affinity and DNase peaks in the chromatin structure of some cell lines. Additional work, including fine-mapping of the GWAS signal to identify the functional variant, is needed to unravel the potential underlying mechanism. Candidate genes could be nearby long intergenic noncoding RNA genes RPH11-526H11-1 and RP11-638L3.1. Increasing evidence suggests that long intergenic noncoding RNAs may play an important role in epigenetic and post-transcriptional regulation in health and (cardiovascular) disease.38,39 However, the characteristics and function of the majority of these RNAs are currently not known. Interrogation of several publicly available databases, such as Genotype-Tissue Expression32 and several long noncoding RNA databases, did not yield additional information. The nearest protein-coding gene, TMX3, lies >500 Kb away but could also be a target given its biological function. Because thioredoxin-related transmembrane protein 3 has been detected at the platelet surface,36 it may play a role in platelet functioning, in line with other members of the protein disulphide isomerase family. Functional follow-up experiments could help to identify and characterize the potential role of these genes in recurrent VT. In addition, long-range chromatin interaction analyses using chromosome conformation capture technologies, such as 4C and Hi-C, might aid to identify other potential target genes.

    Another notable finding is that almost all variants, which have previously been linked to a first VT at genome-wide significance26,27 including the novel risk variants at TSPAN15 and SLC44A2, were not or only nominally associated with the risk of recurrent VT. This is in line with previous reports on the risk variants which have been studied for recurrence risk.36 Several explanations for this discrepancy have been proposed. To some extent, this can be explained by the difference in absolute risks for first and recurrent VT, resulting in the incomparability of effects on a relative risk scale between first and recurrent VT.11 In addition, research into risk factors for recurrence risk may be hindered by index event bias although this could lead to both under- and overestimation of the risk estimate.10 Of note, because all candidate risk variants had effects in the expected direction and 3 of 9 variants were associated with recurrence risk at a significance level of 0.05, which is more than expected by chance, our results provide some evidence that these variants may also impact recurrence risk. In particular, FGG rs2066865 might be promising because earlier studies have also reported some evidence of an association.5,6

    The main limitation of this study is the small sample size with 447 and 345 patients with recurrent VT in the discovery GWAS and the combined replication studies, respectively. As a result, we may have missed associations between recurrent VT and variants with a small effect or a low MAF. The small sample size may also explain why we failed to replicate most suggestively associated variants identified in our GWAS. We, therefore, emphasize the need of a large international collaborative effort to substantially increase the sample size for recurrent VT analyses. Of note, mainly patients of Northwest-European origin were included in our analyses, and, therefore, caution is needed in generalizing our results to other populations. In addition, the X chromosome was not interrogated in the discovery GWAS.

    In both the GWAS and the replication analyses, patients who were lost to follow-up or who experienced an uncertain recurrent VT were considered to be recurrence free. This could have affected our results because we cannot rule out that these patients experienced a recurrent thrombotic event. However, this is unlikely because these patients did not visit the anticoagulation clinics, which monitor anticoagulant treatment. In addition, the results of the sensitivity GWAS, in which these patients were excluded, did not materially differ from the discovery GWAS. Likewise, we obtained a similar effect estimate for rs9946608 in the logistic regression model and the time-to-event analysis, in which patients who were lost to follow-up were censored. Together, this suggests that the impact of misclassification in our study was probably low.

    Our findings could lead to a better understanding of the biological mechanism underlying recurrent VT. In addition, we have previously shown the potential clinical value of genetic risk factors in the risk stratification of first and recurrent VT.5,40 In a preliminary analysis, we showed that adding FV Leiden and rs9946608 to a clinical prediction model slightly improved the risk discrimination of recurrence. Identification of novel risk variants may further improve risk prediction of recurrent VT. Although additional replication and functional analyses are required, we identified a potential risk locus at chromosome 18q22.1 and confirmed the role of FV Leiden in recurrent VT pathophysiology.


    The Data Supplement is available at

    Circ Genom Precis Med is available at

    Correspondence to: Frits R. Rosendaal, MD, PhD, Department of Clinical Epidemiology, Leiden University Medical Center, PO Box 9600, 2300 RC Leiden, The Netherlands. E-mail


    • 1. Prandoni P, Lensing AW, Cogo A, Cuppini S, Villalta S, Carta M, et al. The long-term clinical course of acute deep venous thrombosis.Ann Intern Med. 1996; 125:1–7.CrossrefMedlineGoogle Scholar
    • 2. Hansson PO, Sörbo J, Eriksson H. Recurrent venous thromboembolism after deep vein thrombosis: incidence and risk factors.Arch Intern Med. 2000; 160:769–774.CrossrefMedlineGoogle Scholar
    • 3. Ho WK, Hankey GJ, Quinlan DJ, Eikelboom JW. Risk of recurrent venous thromboembolism in patients with common thrombophilia: a systematic review.Arch Intern Med. 2006; 166:729–736. doi: 10.1001/archinte.166.7.729.CrossrefMedlineGoogle Scholar
    • 4. Marchiori A, Mosena L, Prins MH, Prandoni P. The risk of recurrent venous thromboembolism among heterozygous carriers of factor V Leiden or prothrombin G20210A mutation. A systematic review of prospective studies.Haematologica. 2007; 92:1107–1114.CrossrefMedlineGoogle Scholar
    • 5. van Hylckama Vlieg A, Flinterman LE, Bare LA, Cannegieter SC, Reitsma PH, Arellano AR, et al. Genetic variations associated with recurrent venous thrombosis.Circ Cardiovasc Genet. 2014; 7:806–813. doi: 10.1161/CIRCGENETICS.114.000682.LinkGoogle Scholar
    • 6. Bruzelius M, Ljungqvist M, Bottai M, Bergendal A, Strawbridge RJ, Holmström M, et al. F11 is associated with recurrent VTE in women. A prospective cohort study.Thromb Haemost. 2016; 115:406–414. doi: 10.1160/TH15-06-0459.CrossrefMedlineGoogle Scholar
    • 7. Baglin T, Luddington R, Brown K, Baglin C. Incidence of recurrent venous thromboembolism in relation to clinical and thrombophilic risk factors: prospective cohort study.Lancet. 2003; 362:523–526. doi: 10.1016/S0140-6736(03)14111-6.CrossrefMedlineGoogle Scholar
    • 8. Christiansen SC, Cannegieter SC, Koster T, Vandenbroucke JP, Rosendaal FR. Thrombophilia, clinical factors, and recurrent venous thrombotic events.JAMA. 2005; 293:2352–2361. doi: 10.1001/jama.293.19.2352.CrossrefMedlineGoogle Scholar
    • 9. Prandoni P, Noventa F, Ghirarduzzi A, Pengo V, Bernardi E, Pesavento R, et al. The risk of recurrent venous thromboembolism after discontinuing anticoagulation in patients with acute proximal deep vein thrombosis or pulmonary embolism. A prospective cohort study in 1,626 patients.Haematologica. 2007; 92:199–205.CrossrefMedlineGoogle Scholar
    • 10. Dahabreh IJ, Kent DM. Index event bias as an explanation for the paradoxes of recurrence risk research.JAMA. 2011; 305:822–823. doi: 10.1001/jama.2011.163.CrossrefMedlineGoogle Scholar
    • 11. Cannegieter SC, van Hylckama Vlieg A. Venous thrombosis: understanding the paradoxes of recurrence.J Thromb Haemost. 2013; 11(suppl 1):161–169. doi: 10.1111/jth.12263.CrossrefMedlineGoogle Scholar
    • 12. Zee RY, Bubes V, Shrivastava S, Ridker PM, Glynn RJ. Genetic risk factors in recurrent venous thromboembolism: a multilocus, population-based, prospective approach.Clin Chim Acta. 2009; 402:189–192.CrossrefMedlineGoogle Scholar
    • 13. Jilma B, Kovar FM, Hron G, Endler G, Marsik CL, Eichinger S, et al. Homozygosity in the single nucleotide polymorphism Ser128Arg in the E-selectin gene associated with recurrent venous thromboembolism.Arch Intern Med. 2006; 166:1655–1659. doi: 10.1001/archinte.166.15.1655.CrossrefMedlineGoogle Scholar
    • 14. Mustafa S, Weltermann A, Fritsche R, Marsik C, Wagner O, Kyrle PA, et al. Genetic variation in heme oxygenase 1 (HMOX1) and the risk of recurrent venous thromboembolism.J Vasc Surg. 2008; 47:566–570. doi: 10.1016/j.jvs.2007.09.060.CrossrefMedlineGoogle Scholar
    • 15. Timp JF, Lijfering WM, Flinterman LE, van Hylckama Vlieg A, le Cessie S, Rosendaal FR, et al. Predictive value of factor VIII levels for recurrent venous thrombosis: results from the MEGA follow-up study.J Thromb Haemost. 2015; 13:1823–1832. doi: 10.1111/jth.13113.CrossrefMedlineGoogle Scholar
    • 16. Blom JW, Doggen CJ, Osanto S, Rosendaal FR. Malignancies, prothrombotic mutations, and the risk of venous thrombosis.JAMA. 2005; 293:715–722. doi: 10.1001/jama.293.6.715.CrossrefMedlineGoogle Scholar
    • 17. Aulchenko YS, Ripke S, Isaacs A, van Duijn CM. GenABEL: an R library for genome-wide association analysis.Bioinformatics. 2007; 23:1294–1296. doi: 10.1093/bioinformatics/btm108.CrossrefMedlineGoogle Scholar
    • 18. Howie BN, Donnelly P, Marchini J. A flexible and accurate genotype imputation method for the next generation of genome-wide association studies.PLoS Genet. 2009; 5:e1000529. doi: 10.1371/journal.pgen.1000529.CrossrefMedlineGoogle Scholar
    • 19. Genome of the Netherlands Consortium. Whole-genome sequence variation, population structure and demographic history of the Dutch population.Nat Genet. 2014; 46:818–825. doi: 10.1038/ng.3021.CrossrefMedlineGoogle Scholar
    • 20. Marchini J, Howie B, Myers S, McVean G, Donnelly P. A new multipoint method for genome-wide association studies by imputation of genotypes.Nat Genet. 2007; 39:906–913. doi: 10.1038/ng2088.CrossrefMedlineGoogle Scholar
    • 21. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, Bender D, et al. PLINK: a tool set for whole-genome association and population-based linkage analyses.Am J Hum Genet. 2007; 81:559–575. doi: 10.1086/519795.CrossrefMedlineGoogle Scholar
    • 22. Pruim RJ, Welch RP, Sanna S, Teslovich TM, Chines PS, Gliedt TP, et al. LocusZoom: regional visualization of genome-wide association scan results.Bioinformatics. 2010; 26:2336–2337. doi: 10.1093/bioinformatics/btq419.CrossrefMedlineGoogle Scholar
    • 23. Wang K, Li M, Hakonarson H. ANNOVAR: functional annotation of genetic variants from high-throughput sequencing data.Nucleic Acids Res. 2010; 38:e164. doi: 10.1093/nar/gkq603.CrossrefMedlineGoogle Scholar
    • 24. Devlin B, Roeder K. Genomic control for association studies.Biometrics. 1999; 55:997–1004.CrossrefMedlineGoogle Scholar
    • 25. Ahmad A, Sundquist K, Zöller B, Dahlbäck B, Svensson PJ, Sundquist J, et al. Identification of polymorphisms in a polipoprotein M gene and their relationship with risk of recurrent venous thromboembolism.Thromb Haemost. 2016; 116:432–441. doi: 10.1160/TH16-03-0178.CrossrefMedlineGoogle Scholar
    • 26. Germain M, Chasman DI, de Haan H, Tang W, Lindström S, Weng LC, et al; Cardiogenics Consortium. Meta-analysis of 65,734 individuals identifies TSPAN15 and SLC44A2 as two susceptibility loci for venous thromboembolism.Am J Hum Genet. 2015; 96:532–542. doi: 10.1016/j.ajhg.2015.01.019.CrossrefMedlineGoogle Scholar
    • 27. Hinds DA, Buil A, Ziemek D, Martinez-Perez A, Malik R, Folkersen L, et al; METASTROKE Consortium;INVENT Consortium. Genome-wide association analysis of self-reported events in 6135 individuals and 252 827 controls identifies 8 loci associated with thrombosis.Hum Mol Genet. 2016; 25:1867–1874. doi: 10.1093/hmg/ddw037.CrossrefMedlineGoogle Scholar
    • 28. van Hylckama Vlieg A, Baglin CA, Luddington R, MacDonald S, Rosendaal FR, Baglin TP. The risk of a first and a recurrent venous thrombosis associated with an elevated D-dimer level and an elevated thrombin potential: results of the THE-VTE study.J Thromb Haemost. 2015; 13:1642–1652. doi: 10.1111/jth.13043.CrossrefMedlineGoogle Scholar
    • 29. Johnson AD, Handsaker RE, Pulit SL, Nizzari MM, O’Donnell CJ, de Bakker PI. SNAP: a web-based tool for identification and annotation of proxy SNPs using HapMap.Bioinformatics. 2008; 24:2938–2939. doi: 10.1093/bioinformatics/btn564.CrossrefMedlineGoogle Scholar
    • 30. Willer CJ, Li Y, Abecasis GR. METAL: fast and efficient meta-analysis of genomewide association scans.Bioinformatics. 2010; 26:2190–2191. doi: 10.1093/bioinformatics/btq340.CrossrefMedlineGoogle Scholar
    • 31. Robin X, Turck N, Hainard A, Tiberti N, Lisacek F, Sanchez JC, et al. pROC: an open-source package for R and S+ to analyze and compare ROC curves.BMC Bioinformatics. 2011; 12:77. doi: 10.1186/1471-2105-12-77.CrossrefMedlineGoogle Scholar
    • 32. GTEx Consortium. Human genomics. The Genotype-Tissue Expression (GTEx) pilot analysis: multitissue gene regulation in humans.Science. 2015; 348:648–660. doi: 10.1126/science.1262110.CrossrefMedlineGoogle Scholar
    • 33. Boyle AP, Hong EL, Hariharan M, Cheng Y, Schaub MA, Kasowski M, et al. Annotation of functional variation in personal genomes using RegulomeDB.Genome Res. 2012; 22:1790–1797. doi: 10.1101/gr.137323.112.CrossrefMedlineGoogle Scholar
    • 34. ENCODE Project Consortium. An integrated encyclopedia of DNA elements in the human genome.Nature. 2012; 489:57–74. doi: 10.1038/nature11247.CrossrefMedlineGoogle Scholar
    • 35. Bernstein BE, Stamatoyannopoulos JA, Costello JF, Ren B, Milosavljevic A, Meissner A, et al. The NIH Roadmap Epigenomics Mapping Consortium.Nat Biotechnol. 2010; 28:1045–1048. doi: 10.1038/nbt1010-1045.CrossrefMedlineGoogle Scholar
    • 36. Holbrook LM, Watkins NA, Simmonds AD, Jones CI, Ouwehand WH, Gibbins JM. Platelets release novel thiol isomerase enzymes which are recruited to the cell surface following activation.Br J Haematol. 2010; 148:627–637. doi: 10.1111/j.1365-2141.2009.07994.x.CrossrefMedlineGoogle Scholar
    • 37. Smith NL, Hindorff LA, Heckbert SR, Lemaitre RN, Marciante KD, Rice K, et al. Association of genetic variations with nonfatal venous thrombosis in postmenopausal women.JAMA. 2007; 297:489–498. doi: 10.1001/jama.297.5.489.CrossrefMedlineGoogle Scholar
    • 38. Esteller M. Non-coding RNAs in human disease.Nat Rev Genet. 2011; 12:861–874. doi: 10.1038/nrg3074.CrossrefMedlineGoogle Scholar
    • 39. Uchida S, Dimmeler S. Long noncoding RNAs in cardiovascular diseases.Circ Res. 2015; 116:737–750. doi: 10.1161/CIRCRESAHA.116.302521.LinkGoogle Scholar
    • 40. de Haan HG, Bezemer ID, Doggen CJ, Le Cessie S, Reitsma PH, Arellano AR, et al. Multiple SNP testing improves risk prediction of first venous thrombosis.Blood. 2012; 120:656–663. doi: 10.1182/blood-2011-12-397752.CrossrefMedlineGoogle Scholar


    eLetters should relate to an article recently published in the journal and are not a forum for providing unpublished data. Comments are reviewed for appropriate use of tone and language. Comments are not peer-reviewed. Acceptable comments are posted to the journal website only. Comments are not published in an issue and are not indexed in PubMed. Comments should be no longer than 500 words and will only be posted online. References are limited to 10. Authors of the article cited in the comment will be invited to reply, as appropriate.

    Comments and feedback on AHA/ASA Scientific Statements and Guidelines should be directed to the AHA/ASA Manuscript Oversight Committee via its Correspondence page.